2024· Journal of Lightwave Technology · 42(1)· First author· IF 4.8

End-to-end deep-learning-based photonic-assisted multi-user fiber–mmWave integrated communication system

A. Sun, et al.
Fudan University

TL;DR A 6G-oriented end-to-end deep-learning framework (AMEF) for multi-user photonic-assisted fiber–mmWave integrated communications, comprising a multi-channel model (MCM) and a multi-user transceiver (MUT) with MLP-based encoder/decoder. A two-user 10 km fiber–mmWave experiment delivers 66 Gbps in the wireless-only system and 49.5 Gbps in the integrated fiber-wireless system, with >1.1 dB / >0.6 dB sensitivity gains over mCAP.

Background

6G is moving toward the photonic-assisted mmWave regime, where dynamic spectrum allocation and multi-user transmission place stringent demands on the integrated communication system, while architectures that explicitly handle multi-user spectrum allocation through end-to-end deep learning are still missing.

Framework

We propose AMEF — an Adaptive Multi-user End-to-end Framework — built from a Multi-Channel Model (MCM) that captures the cascaded photonic-assisted fiber + mmWave link, plus a Multi-User Transceiver (MUT) of MLP-based encoder/decoder learning modulation, channel inversion and user separation jointly.

Highlights

Two-user 10 km fiber–mmWave experiment: 66 Gbps wireless and 49.5 Gbps fiber-wireless integrated
>1.1 dB / >0.6 dB receiver-sensitivity gain over mCAP modulation in the two systems
Joint learning of modulation, channel inversion and multi-user separation in one MLP autoencoder pipeline

Citation

A. Sun et al., "End-to-end deep-learning-based photonic-assisted multi-user fiber–mmWave integrated communication system," J. Lightwave Technology, 42(1), 2024.